Question 130
A plugin marketplace adds tool permissions automatically after updates, exposing new write capabilities. Which solution best resolves this operational gap?
- Curate task-specific tools and load infrequent capabilities through controlled discovery.
- Issue a read-scoped credential and remove refund operations from that agent's callable tool set.
- Require version pinning, reviewed capability diffs and explicit authorization on tool upgrades.
- Place an explicit user confirmation and server-side policy check before purchase execution.
Explanation
Correct Answer
C:Dependency upgrades can expand attack surface even when prompt templates are unchanged. Verify governed tool-capability changes under the actual connector identity; the backend must actually enforce governed tool-capability changes, regardless of instructions.
Incorrect Answers
A:Improving smaller authorized tool surface does not imply governed tool-capability changes. The selected design must require version pinning, reviewed capability diffs and explicit authorization on tool upgrades, whereas this choice omits that direct protection or verification step.
B:This addresses read-only tool authority; however, the problem asks for governed tool-capability changes. Without the safeguard to require version pinning, reviewed capability diffs and explicit authorization on tool upgrades, the original failure mode is still possible.
D:This would make sense as an additional project activity for transaction-specific confirmation. The priority required here is governed tool-capability changes, so the workflow still needs a control that will require version pinning, reviewed capability diffs and explicit authorization on tool upgrades.
Question 131
An employee signs into the assistant but the retrieval API uses one unrestricted service account for all document searches. What design decision would best fit these facts?
- Include authorization-aware cache keys and invalidate or recheck permissions when rights change.
- Use delegated identity or a policy enforcement layer that evaluates the caller's actual rights.
- Propagate verified user identity to authorization filtering at retrieval time.
- Reauthorize sensitive operations at execution time and cancel revoked principals' queued work.
Explanation
Correct Answer
C:Login to the chat UI does not confer permission to every document a backend service could read. Exercise document ACL enforcement per user as an actual end user and against forbidden objects, since authenticating a service is not sufficient authorization.
Incorrect Answers
A:The option belongs to the broader work on immediate effect of access revocation, but it cannot by itself deliver document acl enforcement per user. A decision on this case needs the team to propagate verified user identity to authorization filtering at retrieval time and verify that outcome.
B:This shifts attention toward end-user entitlements through service calls when the relevant decision is about document acl enforcement per user. Unless the architecture can propagate verified user identity to authorization filtering at retrieval time, the failure described in the stem remains unresolved.
D:Although execution-time authorization checks can be valuable, it is not the acceptance condition in the question. The missing provision is document acl enforcement per user, addressed by the control to propagate verified user identity to authorization filtering at retrieval time.
Question 132
A user can request another account's invoice by changing an invoice ID in an otherwise legitimate tool call. Which intervention most directly addresses the failure?
- Reauthorize sensitive operations at execution time and cancel revoked principals' queued work.
- Include authorization-aware cache keys and invalidate or recheck permissions when rights change.
- Authorize each requested object against the caller's tenant and ownership.
- Use delegated identity or a policy enforcement layer that evaluates the caller's actual rights.
Explanation
Correct Answer
C:Object-level checks stop cross-account access that endpoint authentication alone would permit. Exercise object-level authorization for tool requests as an actual end user and against forbidden objects, since authenticating a service is not sufficient authorization.
Incorrect Answers
A:That choice would be responsive to execution-time authorization checks, but the reported defect is different. It leaves object-level authorization for tool requests unproved because there is no commitment to authorize each requested object against the caller's tenant and ownership.
B:There is a legitimate use case for immediate effect of access revocation, but it does not supply the test or enforcement needed for object-level authorization for tool requests. The design needs to authorize each requested object against the caller's tenant and ownership.
D:The alternative could reduce risk associated with end-user entitlements through service calls, yet the described behavior requires object-level authorization for tool requests. That outcome depends on the capability to authorize each requested object against the caller's tenant and ownership.
Question 133
A remote MCP connector accepts bearer tokens but never checks their audience or expiration. What should the solution architect prioritize?
- Validate issuer, audience, lifetime and scopes before granting connector access.
- Use delegated identity or a policy enforcement layer that evaluates the caller's actual rights.
- Reauthorize sensitive operations at execution time and cancel revoked principals' queued work.
- Keep credentials in an encrypted server-side vault and send only opaque operation handles to the model.
- Diagnose the missing API scope and resource policy, then grant the minimum required action.
Explanation
Correct Answer
A:Token possession is insufficient if the service fails to confirm the token was minted for that resource and operation.
Incorrect Answers
B:By itself, this improves end-user entitlements through service calls, but the case is not asking the architect to optimize that dimension. The requirement is proper credential validation, supported by an approach to validate issuer, audience, lifetime and scopes before granting connector access.
C:Its main effect concerns execution-time authorization checks. The case explicitly hinges on proper credential validation; to resolve it reliably, the implementation has to validate issuer, audience, lifetime and scopes before granting connector access, which this choice does not establish.
D:The option might improve secrets separation from model context; that remains insufficient for proper credential validation. The case is resolved by proper credential validation, which depends on a design that will validate issuer, audience, lifetime and scopes before granting connector access.
E:The proposed action prioritizes minimum-scope remediation for api denial. That can coexist with the right solution, but it cannot replace proper credential validation: the application must still validate issuer, audience, lifetime and scopes before granting connector access.
Question 134
A third-party integration stores customer OAuth refresh tokens inside prompt history for convenience. Which architecture adjustment is most appropriate?
- Include authorization-aware cache keys and invalidate or recheck permissions when rights change.
- Reauthorize sensitive operations at execution time and cancel revoked principals' queued work.
- Keep credentials in an encrypted server-side vault and send only opaque operation handles to the model.
- Validate issuer, audience, lifetime and scopes before granting connector access.
Explanation
Correct Answer
C:Prompt content is not a secrets store; credential rotation and revocation require controlled storage. Exercise secrets separation from model context as an actual end user and against forbidden objects, since authenticating a service is not sufficient authorization.
Incorrect Answers
A:Although immediate effect of access revocation can be valuable, it is not the acceptance condition in the question. The missing provision is secrets separation from model context, addressed by the control to keep credentials in an encrypted server-side vault and send only opaque operation handles to the model.
B:It is aimed at a different outcome, execution-time authorization checks. No evidence is given that it provides secrets separation from model context; the needed mechanism is to keep credentials in an encrypted server-side vault and send only opaque operation handles to the model at the appropriate boundary.
D:That is primarily a technique for proper credential validation. It lacks the explicit measure for secrets separation from model context that is necessary here, namely a system that will keep credentials in an encrypted server-side vault and send only opaque operation handles to the model.
Question 135
An agent service authenticates to an API as itself but needs to preserve the end user's document entitlements. Which next step is most defensible?
- Propagate verified user identity to authorization filtering at retrieval time.
- Reauthorize sensitive operations at execution time and cancel revoked principals' queued work.
- Authorize each requested object against the caller's tenant and ownership.
- Use delegated identity or a policy enforcement layer that evaluates the caller's actual rights.
Explanation
Correct Answer
D:Application authentication and end-user authorization are separate requirements in a multi-user assistant. Exercise end-user entitlements through service calls as an actual end user and against forbidden objects, since authenticating a service is not sufficient authorization.
Incorrect Answers
A:The alternative could reduce risk associated with document acl enforcement per user, yet the described behavior requires end-user entitlements through service calls. That outcome depends on the capability to use delegated identity or a policy enforcement layer that evaluates the caller's actual rights.
B:The proposed measure is aimed at execution-time authorization checks, not end-user entitlements through service calls. The architecture must instead use delegated identity or a policy enforcement layer that evaluates the caller's actual rights; a separate change is still needed for end-user entitlements through service calls.
C:The relevant trade-off is not primarily object-level authorization for tool requests. A successful fix needs end-user entitlements through service calls; otherwise end-user entitlements through service calls remains unverified even after this change.
Question 136
A tenant changes an employee's privileges but cached RAG results remain accessible for the rest of the day. Which approach should be recommended?
- Use delegated identity or a policy enforcement layer that evaluates the caller's actual rights.
- Include authorization-aware cache keys and invalidate or recheck permissions when rights change.
- Log the employee’s verified identity on each query but continue serving cache entries until their old TTL expires.
- Authorize each requested object against the caller's tenant and ownership.
Explanation
Correct Answer
B:Previously cached evidence must not outlive the caller's access entitlement. Exercise immediate effect of access revocation as an actual end user and against forbidden objects, since authenticating a service is not sufficient authorization.
Incorrect Answers
A:The option might improve end-user entitlements through service calls; that remains insufficient for immediate effect of access revocation. The case is resolved by immediate effect of access revocation, which depends on a design that will include authorization-aware cache keys and invalidate or recheck permissions when rights change.
C:Identity logging can make unauthorized cache reads visible after the fact, but it does not prevent them. The stale authorization decision remains in use until expiry, contrary to the requirement for immediate privilege revocation.
D:This addresses object-level authorization for tool requests; however, the problem asks for immediate effect of access revocation. Without the safeguard to include authorization-aware cache keys and invalidate or recheck permissions when rights change, the original failure mode is still possible.
Question 137
A developer grants broad cloud administrator access to fix a 403 from a single CRM integration. What is the best production design response?
- Diagnose the missing API scope and resource policy, then grant the minimum required action.
- Validate issuer, audience, lifetime and scopes before granting connector access.
- Use delegated identity or a policy enforcement layer that evaluates the caller's actual rights.
- Propagate verified user identity to authorization filtering at retrieval time.
Explanation
Correct Answer
A:A forbidden response needs a targeted permission trace, not blanket privilege escalation. Exercise minimum-scope remediation for API denial as an actual end user and against forbidden objects, since authenticating a service is not sufficient authorization.
Incorrect Answers
B:Improving proper credential validation does not imply minimum-scope remediation for api denial. The selected design must diagnose the missing API scope and resource policy, then grant the minimum required action, whereas this choice omits that direct protection or verification step.
C:This recommendation targets end-user entitlements through service calls at a different point in the workflow. It leaves minimum-scope remediation for api denial outside the design contract and does not require the team to diagnose the missing API scope and resource policy, then grant the minimum required action.
D:This is a valid architectural concern when optimizing document acl enforcement per user. Here it leaves minimum-scope remediation for api denial unaddressed, whereas the intended change must diagnose the missing API scope and resource policy, then grant the minimum required action before the case can be considered resolved.
Question 138
A sensitive queued job sometimes executes after the requesting employee loses access. Separately, an OAuth connector callback does not validate authorization-request state or proof of key possession, creating a code-substitution risk. Which TWO controls close these distinct authorization gaps? Choose TWO.
- Authorize each requested object against the caller's tenant and ownership.
- Bind the OAuth redirect to its initiating session with validated state and PKCE during authorization-code exchange.
- Reauthorize sensitive operations at execution time and cancel revoked principals' queued work.
- Propagate verified user identity to authorization filtering at retrieval time.
- Use delegated identity or a policy enforcement layer that evaluates the caller's actual rights.
Explanation
Correct Answers
B:Checking state protects the login transaction against cross-request substitution, while PKCE binds the exchanged code to the initiating client. Token-scope checking alone does not establish that the callback belongs to the original consent request.
C:Submitting a job does not preserve indefinite authority to execute it. Sensitive queued actions need a fresh check of employee status and resource permissions at execution, with revoked requests cancelled before any side effects occur.
Incorrect Answers
A:The control would be justified if object-level authorization for tool requests were the limiting factor. In this scenario the decisive target is execution-time authorization checks and oauth callback binding using state and pkce; success requires an architecture that will bind the OAuth redirect to its initiating session with validated state and PKCE during authorization-code exchange.
D:For a workflow whose primary weakness was document acl enforcement per user, that proposal could be reasonable. Here the unmet obligation is execution-time authorization checks and oauth callback binding using state and pkce, and that requires the system to reauthorize sensitive operations at execution time and cancel revoked principals' queued work.
E:Although end-user entitlements through service calls can be valuable, it is not the acceptance condition in the question. The missing provision is execution-time authorization checks and oauth callback binding using state and pkce, addressed by the control to bind the OAuth redirect to its initiating session with validated state and PKCE during authorization-code exchange.
Question 139
An MCP gateway trusts the client application's label for which organization owns the request. Which solution best resolves this operational gap?
- Authorize each requested object against the caller's tenant and ownership.
- Use delegated identity or a policy enforcement layer that evaluates the caller's actual rights.
- Propagate verified user identity to authorization filtering at retrieval time.
- Derive tenant identity from verified credentials and enforce it at the gateway.
Explanation
Correct Answer
D:Client-supplied metadata can be forged; tenancy boundaries need trusted identity assertions. Exercise verified tenant binding at connector boundary as an actual end user and against forbidden objects, since authenticating a service is not sufficient authorization.
Incorrect Answers
A:The proposed measure is aimed at object-level authorization for tool requests, not verified tenant binding at connector boundary. The architecture must instead derive tenant identity from verified credentials and enforce it at the gateway; a separate change is still needed for verified tenant binding at connector boundary.
B:That choice would be responsive to end-user entitlements through service calls, but the reported defect is different. It leaves verified tenant binding at connector boundary unproved because there is no commitment to derive tenant identity from verified credentials and enforce it at the gateway.
C:There is a legitimate use case for document acl enforcement per user, but it does not supply the test or enforcement needed for verified tenant binding at connector boundary. The design needs to derive tenant identity from verified credentials and enforce it at the gateway.
Question 140
An interactive support assistant makes three serial retrieval calls although all evidence queries are independent. What design decision would best fit these facts?
- Cache authorized stable retrieval results with version and TTL validation.
- Adopt separate fast and deep-service paths with explicit deadlines and fallback behavior.
- Run independent reads concurrently and merge results with bounded timeouts.
- Use exponential backoff with jitter and a bounded retry budget for transient failures.
- Run accuracy-versus-latency evaluations across representative hard and easy queries.
Explanation
Correct Answer
C:Parallel reads reduce critical-path latency when results have no dependency on one another. Measure concurrent evidence retrieval under representative concurrency and source freshness constraints; low latency is not success when evidence for concurrent evidence retrieval is missing.
Incorrect Answers
A:This is a valid architectural concern when optimizing safe caching of stable evidence. Here it leaves concurrent evidence retrieval unaddressed, whereas the intended change must run independent reads concurrently and merge results with bounded timeouts before the case can be considered resolved.
B:This would make sense as an additional project activity for quality-aware deadline management. The priority required here is concurrent evidence retrieval, so the workflow still needs a control that will run independent reads concurrently and merge results with bounded timeouts.
D:Its main effect concerns throttling-aware retry policy. The case explicitly hinges on concurrent evidence retrieval; to resolve it reliably, the implementation has to run independent reads concurrently and merge results with bounded timeouts, which this choice does not establish.
E:The option might improve validated speed and recall tradeoff; that remains insufficient for concurrent evidence retrieval. The case is resolved by concurrent evidence retrieval, which depends on a design that will run independent reads concurrently and merge results with bounded timeouts.